Hi, while implementing cpupool support in XEN I had to track down some issues with deadlocks. Is there any interest in patches for the following? - compile time switchable spinlock timeout detection - printout of state of ALL cpus to console in case of hypervisor crash - some enhancements in lock handling to reduce (potential) latency: Today in *lock_irq and *lock_irqsave interrupts are disabled and then spinning is started. If the lock is not obtained, enabling and disabling of IRQs in the spinloop would block IRQs as short as possible. Juergen -- Juergen Gross Principal Developer IP SW OS6 Telephone: +49 (0) 89 636 47950 Fujitsu Siemens Computers e-mail: juergen.gross@fujitsu-siemens.com Otto-Hahn-Ring 6 Internet: www.fujitsu-siemens.com D-81739 Muenchen Company details: www.fujitsu-siemens.com/imprint.html _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
On 25/03/2009 08:24, "Juergen Gross" <juergen.gross@fujitsu-siemens.com> wrote:> while implementing cpupool support in XEN I had to track down some issues > with deadlocks. Is there any interest in patches for the following? > > - compile time switchable spinlock timeout detectionNMI watchdog support effectively gets us this, and more general deadlock detection. Just put ''watchdog'' on Xen''s command line.> - printout of state of ALL cpus to console in case of hypervisor crashNot sure how useful this would be really for most crashes. Probably we don''t want it.> - some enhancements in lock handling to reduce (potential) latency: > Today in *lock_irq and *lock_irqsave interrupts are disabled and then > spinning is started. If the lock is not obtained, enabling and disabling of > IRQs in the spinloop would block IRQs as short as possible.Could do. I don''t think our IRQ locks are very contended, but equally I suppose the patch is only a few lines. -- Keir _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Keir Fraser wrote:> On 25/03/2009 08:24, "Juergen Gross" <juergen.gross@fujitsu-siemens.com> > wrote: > >> while implementing cpupool support in XEN I had to track down some issues >> with deadlocks. Is there any interest in patches for the following? >> >> - compile time switchable spinlock timeout detection > > NMI watchdog support effectively gets us this, and more general deadlock > detection. Just put ''watchdog'' on Xen''s command line.Sigh. I should have read ALL the code ;-)> >> - printout of state of ALL cpus to console in case of hypervisor crash > > Not sure how useful this would be really for most crashes. Probably we don''t > want it.Okay. I''ll let it rot here. :-)> >> - some enhancements in lock handling to reduce (potential) latency: >> Today in *lock_irq and *lock_irqsave interrupts are disabled and then >> spinning is started. If the lock is not obtained, enabling and disabling of >> IRQs in the spinloop would block IRQs as short as possible. > > Could do. I don''t think our IRQ locks are very contended, but equally I > suppose the patch is only a few lines.I''ll prepare a patch and let you decide. Juergen -- Juergen Gross Principal Developer IP SW OS6 Telephone: +49 (0) 89 636 47950 Fujitsu Siemens Computers e-mail: juergen.gross@fujitsu-siemens.com Otto-Hahn-Ring 6 Internet: www.fujitsu-siemens.com D-81739 Muenchen Company details: www.fujitsu-siemens.com/imprint.html _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel